I am simulating a 3D airfoil at higher angle of attacks (post stall regime) . I wanted to know that is RANS formulation capable enough to resolve tip vertices or not .
From a theoretical point of view, the analysis makes sense using URANS if the simulation has the aim to solve the flow at a range of time-varying AOA (just think about the similar URANS application in the in-cylinder flow with a moving piston). That means that at each AOA you compute a statistical solution for each possible realization.
What is theoretically flawed is the URANS at a fixed AOA with the aim of getting a time-dependent solution. It is quite debatable the meaning of such a solution.
As addressed by Tapan K. Sengupta it is instructive to explore this issue.
RANS is a statistically steady formulation, thus - by definition - you can get only a smooth solution representing the zero-th statistics and only averaged vortical regions are in the solution. At high angle of attack I suspect you can not get a convergent steady solution.
Differently, you can consider unsteady approaches such as URANS (with the limits in its real meaning), DES or LES going in increasing order of complexity.
Filippo Maria Denaro Thanks for your response . I was wondering would RANS with a suitable turbulence model solve this problem ? I dont have that much computing power to run DES or LES simulations .
Filippo Maria Denaro At high angle of attacks due to the asymmetry of wing tip vertices we witness a phenomenon called wing rock . I want to study this phenomenon using self imposed lateral oscillations That is my goal .
Yes I am aware of the RANS formulation , the fluctuations in the results are averaged out to give a steady state mean flow solution .
See here sec.9.6 http://www.dept.aoe.vt.edu/~mason/Mason_f/ConfigAeroHiAlphaNotes.pdf
the description of a non-symmetric flow solution even for symmetric body is typical of an unsteady flow behavior.
Despite of any turbulence model you could adopt, the RANS solution cannot produce such bifurcation in the solution. Think about the vortex shedding, the RANS solution will provide always two steady vortical regions.
You can only try to convince yourself by performing some RANS test (if you get convergence to a steady state...). Swithing to URANS could give some solution mimicking the flow structures, in the limits of the URANS meaning.
It is an interesting proposition. As Prof. Denaro notes that you will get a time averaged field. If finite wing theory has any relevance with real flow over the wing, then your RANS solution will show some closeness with Prandtl-Lancaster model. However, that is a model for steady state.
If instead, you are interested in tip vortex shedding problem, then be prepared for surprises. I would not think that URANS will capture the correct time scale and spatial scales of the event. Nonetheless, it is worthwhile to pursue it and report the results. Even if it turns out to be negative!
Please see the papers links provided below. They have discussed the accuracy and reliability of a RANS model (k-omega SST) in predicting the tip vortex phenomenon of 3-D configuration. I guess the same authors have published other works on the same subject.
Article Analysis of tip vortex inception prediction methods
Article Evaluation of curvature correction methods for tip vortex pr...
From a theoretical point of view, the analysis makes sense using URANS if the simulation has the aim to solve the flow at a range of time-varying AOA (just think about the similar URANS application in the in-cylinder flow with a moving piston). That means that at each AOA you compute a statistical solution for each possible realization.
What is theoretically flawed is the URANS at a fixed AOA with the aim of getting a time-dependent solution. It is quite debatable the meaning of such a solution.
As addressed by Tapan K. Sengupta it is instructive to explore this issue.
I would request Prof. Filippo Maria Denaro to spell out the theoretical flaw of URANS for fixed AOA cases producing time dependent solution. This is very important, as at any point in time, there must be thousands of researchers precisely doing the same!! They should be alerted about the pitfalls.
I would highlight that we can find in literature many studies using URANS and showing solutions mimicking the typical LES behavior. As a counterpart, who experienced RANS solutions that did not succeed in convergence towards a steady state, they can report that the oscillating fields "resemble" a transient URANS solution...Thus, what about such a confusing framework about RANS/URANS/LES solutions??? What is the theoretical basis for saying that a statistical formulation can produce physical and relevant time dependent solutions without an external time depending forcing (e.g. a unique steady AOA)?
Filippo Maria Denaro , Tapan K. Sengupta In my case the steady "RANS formulation results" are converging to one point (a single value) . However , they have a significant error when are compared with experimental results . Here the results I am talking about are aerodynamic coefficients .Lift is well predicted and error is below 10 % when compared with experimental values . Drag is about 20 to 25 % under-predicted by the RANS .
I have used RANS even at angle of attacks as high as 18 Degree and 20 Degree. I did not shift to URANS because RANS results were converging to a steady state solution .
I am not sure how URANS or LES would effect the results .
One can get lift values, even using panel method, without the need to solve Navier-Stokes equation. Drag value and unsteadiness are two hall-marks for solving Navier-Stokes equation. Obviously RANS fails their badly. I am not surprised. Next, can you try URANS?
Also, are you using a rectangular wing planform? If so, then check out the basis of finite wing theory due to Prandtl-Lancaster. I have seen some researchers even try using finite wing theory for flapping wing aerodynamics. While those may be too adventurous, is it time to red-flag such resutls? You can try to check it out.
Tapan K. Sengupta I am running URANS now and yes I am using a rectangular wing planform . I will look into finite wing theory as per your recommendation.
Can you mention some literature regarding RANS not good at predicting the drag values accurately ?
I never use open-source software for our research, even though we have used SU2 for developmental work. So, I do not have ready answer for you. Prof. Filippo Maria Denaro will be able to provide you such details. He is very conversant with these issues.
I will wait for your results on finite wing theory.
The problems in predicting drag are not limited to RANS but extended to URANS, LES, DES... the key is that the viscous drag depends on the stress at the wall and we need to compute accurately this value on a sufficiently fine grid such that the numerical derivative is well resolved. The flow close to the wall is practically laminar (viscous sub-layer) but in a layer of very low thickness. That means to have a grid with at least 3-4 points within y+
Filippo Maria Denaro While making mesh I took good care of the y+ values and first cell height . I kept the y+ value equal to 1 . Do you think I need to decrease it further .
Plus if it is the issue of viscous drag due to stresses at the wall , How come I am getting accurate values of drag at 2 and 4 degree angle of attacks ?